Skip to main content

Guides

How Do AI Text Detectors Work?

CleanPastedText editorial · Updated

Text workbench

Try it on your text

Cleaning mode

Removes hidden characters and standardizes AI-style punctuation.

Try a real example

Each sample contains a problem you cannot see.

Original

Pasted text

0 chars · 0 words

Cleaned

Ready to copy

0 changes

Text stays in this browser

Same words · No AI rewriting · No content logging

What signal does an AI detector actually read?

An AI text detector never sees "AI" or "human" as a label in the data — it estimates a probability from two families of signal in the writing itself: how predictable the words are, and how much that predictability changes across the document.

Perplexity

How predictable a piece of text is to a reference language model. When a model can easily guess the next word, perplexity is low. Language models generate by picking statistically likely next tokens, so AI text tends to score consistently low.

Burstiness

How much that predictability varies from sentence to sentence across a document. Human writers vary sentence length and word choice unevenly; models apply similar next-word logic throughout, so their burstiness stays flat.

Why did GPTZero move past perplexity and burstiness?

Perplexity and burstiness are easy to explain but easy to game — paraphrasing tools and "humanizers" exist specifically to push those two numbers toward human-looking ranges. GPTZero says it stopped relying on perplexity and burstiness alone in autumn 2023 and moved to a deep-learning classifier architecture, with perplexity and burstiness retained as just one of several indicators — alongside novel-text search and other model-based signals — rather than the whole verdict. Other detectors, including ZeroGPT, Copyleaks, and Originality.ai, still lean more heavily on perplexity- and burstiness-style scoring.

How accurate are AI detectors, really?

Vendor claims and independent testing tell noticeably different stories, and the gap matters most for anyone facing a consequence — a student, a job applicant, a freelance writer.

ClaimSourceWhat it found
Vendor accuracy claimTurnitin's July 2026 model updateUnder 1% false-positive rate claimed; scores in the 1%-19% range are suppressed and not shown, specifically to avoid flagging borderline text.
Independent testingStanford-linked research (Liang et al.)4-9% false positives on human-written text in broader testing; on TOEFL essays by non-native English writers, over half were misclassified as AI-generated.
Character-level testsOriginality.aiAdding or removing invisible Unicode characters left detection scores essentially unchanged — detectors read words, not bytes.

The bias finding is the sharpest of the three. In "GPT detectors are biased against non-native English writers" (Liang, Yuksekgonul, Mao, Wu & Zou, published in Patterns, 2023), the authors ran seven widely used detectors against real TOEFL essays and found near-perfect accuracy on native-speaker eighth-grade essays but more than 50% of non-native essays flagged as AI-generated. Their hypothesis: non-native writing tends to have less lexical variety, which lowers perplexity — the exact signal several detectors read as evidence of AI authorship.

What a detector score does — and doesn't — tell you

Even Turnitin, the detector making the strongest accuracy claim, tells institutions not to use its score as the sole basis for an academic-integrity decision. Treat any AI-detection percentage as one input to review, not a verdict — and treat any tool that promises to help you "beat" a detector as making a claim it cannot back up. Detectors score word patterns; they do not read hidden characters, and statistical watermarks like SynthID-Text are verified with a secret key the public tools don't have, not estimated from writing style at all.

This site's own scope is narrower and more honest than a detector or an evasion tool: the AI text cleaner finds and removes character-level artifacts — the kind ChatGPT briefly embedded in 2025 — and shows you exactly what it found with the AI watermark checker. It does not, and cannot, change your detector score.

Common questions

Frequently asked questions

How do AI text detectors like GPTZero and Turnitin actually work?

Most start from two signals: perplexity (how predictable each word is to a reference language model — AI text tends to pick likely words, so perplexity is low) and burstiness (how much that predictability swings between sentences — human writing swings more). Modern tools layer a trained deep-learning classifier and, in GPTZero's case, a search against known AI outputs on top of those signals rather than relying on perplexity and burstiness alone.

How accurate are AI detectors in 2026?

It depends heavily on the tool and the text. Turnitin says its July 2026 model update holds false positives under 1% and suppresses any score or highlight in the 1%-19% range specifically to avoid flagging borderline cases. Independent researchers have measured higher error rates on real-world text — a Stanford-linked study found detectors misclassifying more than half of non-native-English TOEFL essays as AI-written, and other independent testing has found 4-9% false positives on human writing generally.

Can removing hidden Unicode characters fool an AI detector?

No. AI detectors score word choice and sentence statistics, not invisible characters. Originality.ai's own testing found that adding or removing invisible characters left detection scores essentially unchanged. Stripping characters with a tool like this one is about formatting hygiene and privacy, not detector evasion — treat any tool that claims otherwise with suspicion.

Do AI detectors catch statistical watermarks like SynthID-Text?

Not directly, and that is the point of the distinction. A statistical watermark (Google's SynthID-Text, or Anthropic's watermark on Claude models from August 2026) is verified with a secret key held by the model vendor, not scored by a general-purpose detector reading the text cold. A public detector like Turnitin or GPTZero is estimating AI authorship from writing patterns alone, without that key.

Why do AI detectors flag non-native English writers more often?

A 2023 Stanford-led study (Liang et al., published in Patterns) found GPT detectors misclassified more than half of TOEFL essays written by non-native English speakers as AI-generated, versus near-perfect accuracy on US eighth-grade essays written by native speakers. The likely cause: non-native writing often has lower lexical variety, which lowers perplexity — the same signal detectors read as a sign of AI generation.

Is a high AI-detection score proof a text was written by AI?

No. Every major detector, including Turnitin, describes its score as a directional signal to review, not standalone proof, and Turnitin explicitly advises against using its score as the sole basis for an academic-integrity decision. Scores are probabilities estimated from writing patterns; they can be wrong in both directions.

Continue reading

Related guides & tools